1.3.2: Flexible Dashboard Time Ranges & a Stable Geo Map

A feature release focused on time ranges in custom dashboards, driven by two reports from @SemoTech running LogTide against a real WAN edge. Picking a window is now consistent across every panel type and reaches the common ops ranges that were missing, a new board-level picker retimes a whole dashboard in one move, and the geo map stops throwing away your framing every time it refreshes. Two panels turned out to be ignoring their configured window entirely, which is fixed here.

Added

• One time range list for every panel (#305): panel windows used to be a small, inconsistent set that differed by panel type, with no way to express common ops ranges like 48 hours or a few days without jumping straight to a week. Every window-scoped panel now shares the same list: 15 minutes, 1 hour, 6 hours, 12 hours, 24 hours, 48 hours, 3 days, 7 days, 14 days and 30 days. That covers time_series, top_n_table, geo_map, log_table, metric_chart, metric_stat, trace_latency, trace_volume, detection_events and activity_overview. The list is defined once and shared by the API validator and every config form, so the two can never drift apart, and it is a strict superset of what each panel accepted before: existing dashboards and YAML exports keep working untouched, with no migration and no schema version bump. Config forms now label the control “Time range” everywhere and spell the values out in full, so you pick “Last 48 hours” rather than decoding 48h
• Dashboard-wide time picker (#305): a new control in the dashboard toolbar retimes every window-scoped panel at once, which is what you want during an ops review when the question is “show me this whole board over the last three days”. It is deliberately a view-level control: your per-panel settings and the saved dashboard are never modified, live log tables and panels without a window are left alone, and choosing “Panel defaults” hands every panel back its own configured range. Because nothing is persisted, a shared dashboard always opens the way its author set it up

Fixed

• Time series panels ignored their window (#305): the time_series panel always rendered the last 24 hours no matter which interval was configured, and silently dropped its service filter as well, so a panel set to 7 days and one service showed 24 hours of everything. Both are fixed. The panel now picks a bucket size to match the window (from one-minute buckets on a 1-hour view up to daily buckets on a 30-day one), fills empty buckets with zeros so a quiet period reads as quiet instead of vanishing from the chart, and switches the axis to dates once the range spans more than a couple of days
• Top services tables ignored their window (#305): the top_n_table panel’s service dimension always aggregated the last 7 days regardless of the configured interval, so a “Top services (24h)” panel was really showing a week. It now honors the window, while keeping the fast pre-aggregated path for the 7-day default so the common case stays as quick as before. The Last seen column, previously unavailable on this dimension, now works here too
• Geo map viewport reset on every refresh (#304): the map was torn down and rebuilt from scratch on each poll, which re-centered and re-zoomed it and made a dashboard left open on screen visibly jump every refresh cycle. Any pan or zoom you had done was lost. The map instance now persists and only its data points are swapped, so your framing survives background refreshes. Automatic fitting happens where it makes sense, on first paint and when you change what the panel queries, and a new Fit data control re-frames on demand. Two related annoyances went with it: the mouse wheel over the panel now scrolls the dashboard instead of zooming the map, so scrolling past a map no longer destroys its framing, and the map no longer paints over open dialogs